BronxBomber Tribune

Due to my fault in data mining, I divided posts into two.
In Part I, I'll talk about pitch classification, and the properties of each repertoires based on Pitch F/X.

I followed the methodology of Derek Carty on The Hardball Times - About Cliff Lee Part II.

- Sample data : I got 1,246 pitches and 1,197 balls had pitchF/X info.



1. Pitch classification

As far as officialy announced, Tim Lincecum throws
1. Fastball + Curve + Changeup(by Josh Kalk 2007)
2. Fastball + Curve + Changeup + Hard sliderby BA 2007)
these 3 or 4 repertoires. I'll follow the Josh way to classify Lincecum's repertoires.

- Raw Movement DATA

This is the raw state(does not have pitch_type information) movement data of Lincecum's every pitch between 4/1/08 and 6/4/08.
1) Classification by horizontal movement
As you see, there's two different groups divided by horizontal movement. We can divide these two by criteria of 90 MPH, one will be fastball+expected hard slider group, the other will be curveball+changeup group.

2) Classification by vertical movement
We can divide 3(or 4) groups by vertical movement.
Based on vertical movement, we can assume that Lincecum did not threw 'Hard slider'.

- 1st Pitch Classification

I classified Raw datas into 1st classified datas by horizontal - vertical movement.

See above, we have some 85+MPH curveballs. The problem with these curveballs are,
1. Threre are curveballs which have vertical movement value over 0.
2. There is some curveball which has the speed near 90MPH.
And, we can see some curveballs that have pfx_z value below zero. I think these should be inserted into changeup, not curveballs.

So, I classified 1st datas by conditions below,
1. If curveball has value of pfx_x below 2, change these to changeups.
2. All balls over 87MPH should be converted to fastballs.
3. Fastballs which has below 87MPH in speed and pfx_x >2, then convert to curveballs.
4. Fastballs which has below 87MPH in speed and pfx_x <2, then convert to changeups.
5. If we have other vague datas even finished 4 procedures above, remove these datas within resonable range.

- Final Pitch Classification

As you can see, we classified 3 pitches.


2. Fastball

- 2-seamer or 4-seamer?

Lincecum's fastball has big range of horizontal movement values on his fastballs. That is, Lincecum can throw balls just simillar as straight 4-seamer, but he can even throw fastballs containing wicked horizontal movements. Plus 94.5 MPH in average speed.
What do you feel, if you have to face wicked moving fastball over 95MPH?
(I don't even want to think of it.)

- Divide fastballs by -3.87 pfx_x value

The average value of pfx_x was -3.87. So I serarated fastballs by 2 groups, one is FA -3.87, the other is FA +3.87(fastballs that has pfx_x value over -3.87).
The average property values of these two was,
FA -3.87 : 95.27MPH, pfx_x -7.11, pfx_z 12.21 (356 pitches)
FA +3.87 : 93.94MPH, pfx_x -1.19, pfx_z 10.34 (431 pitches)

- Release Point


As you see above, I cannot get any special results over release point distribuion. But, we can think when Lincecum throws FA -3.87 kind of balls, the release point would be slightly right compared to FA +3.87 kind of balls.

- Side View


FA +3.87 starts higher and reaches lower than FA -3.87 balls.
It means FA +3.87 has more sinking power than FA -3.87 balls.

- Top View


FA +3.87 has just straight trajectory, like four-seam fastball.
But for FA -3.87, it has some significant change in horizontal movement. Lincecum's FA -3.87 fastball has a sort of "Boomerang" effect.(when I separated 2 groups by pfx_x -6.00, the result was same.) We already know that TV broadcasts distort original trajectory of balls, because they do not set their cameras on front position. Lincecum's FA -3.87 fastball really changes its horizontal moving direction.

Mysterious fastball : FA -3.87

But, if we think FA -3.87 as two-seam fastball and FA +3.87 as four-seam fastball, there's a problem.
Do two-seam fastballs have higher speed than four-seam fastballs in general?
As far as I have known, four-seam fastballs are faster than two-seam fastballs.
Then what? FA -3.87 is faster than FA +3.87 - two-seam fastballs are faster than four-seam fastballs? or FA +3.87 is a sort of slider?

There's big physical problems for regarding FA +3.87 as sliders.
1) FA +3.87 does not satisfy the trajectory of average sliders.
2) FA +3.87 has more pfx_z than average sliders in league.
3) and FA +3.87 has average speed over 93MPH.


3. Curveballs

Slow curve and Power curve?

I separated curveballs into 2 groupes, by speed of 80 MPH - CU -80, CU +80.
CU -80 : 77.66 MPH, 6.11 pfx_x, -5.35 pfx_z(65 pitches) - I thought this would satisfy normal curveball.
CU +80 : 82.50 MPH, 5.69 pfx_x, -3.77 pfx_z(74 pitches) - and this would be power curve.

Release Point


Nothing special, but Lincecum has almost same release point distribution on both curveballs and fastballs.

Side View


CU +80 and -80 has very different trajectories. CU +80 has splitter-like trajectory, but CU -80 has somewhat weak movement compared to average curveballs.

Top View

CU +80 bends toward LHB, CU -80 has 'hook' power on it.

Overall
CU -80 kind is the slowest balls thrown by Lincecum and has slow-hook like trajectory.
CU +80 kind is 5 MPH faster than CU -80 kind(in average start speed), and has splitter or power-curve like trajectory.


4. Changeup

Average value of properties
82.83 MPH, -3.85 in pfx_x, 5.28 in pfx_z(216 pitches)

Release Point

Lincecum throws all of his pitches on almost exact same point.(In other words, Lincecum's release point distribution has no big difference in each pitches.)
Thus,
"Batters cannot judge what Lincecum has pitched by his release point."

Side View

Blue is the trajectory of changeup. Changeups are similar to CU +80 balls in average start speed, but changeups have almost straight trajectory.

Top View

As you see, changeups and CU +80 curveballs have mirror relationship in their horizontal movements.

So I dare say, Lincecum's changeup is more powerful when combined with his powercurves. Because they are same in start speed, but very different in movements.


5. See you again!

These are all for today. Thank you for reading, and any comments/questions are welcomed.
If you have any problems with writing comment on my blog, just send me a e-mail.
landor82 at gmail.com.

special thanks to : Josh Kalk, Mike fast, Derak Carty, Jonathan Hale, Harry Pavlidis and Ike hall.